Skip to content

Comments

Support TensorRT-LLM Uses FlexKV in Multi-node Tensor Parallelism#53

Merged
axxx03 merged 7 commits intotaco-project:devfrom
axxx03:feature/trtllm_adapter_tp16
Nov 27, 2025
Merged

Support TensorRT-LLM Uses FlexKV in Multi-node Tensor Parallelism#53
axxx03 merged 7 commits intotaco-project:devfrom
axxx03:feature/trtllm_adapter_tp16

Conversation

@axxx03
Copy link
Collaborator

@axxx03 axxx03 commented Nov 20, 2025

No description provided.

charliecgxu pushed a commit to charliecgxu/FlexKV that referenced this pull request Nov 24, 2025
@axxx03 axxx03 force-pushed the feature/trtllm_adapter_tp16 branch from d88efa8 to ff9f10b Compare November 25, 2025 07:14
MAX_NUM_TOKENS=16384
# MAX_SEQ_LEN=8192
# MAX_NUM_TOKENS=8192
HOSTFILE=/cfs_zhongwei/rongwei/scripts/flexkv/trtllm_hostfile
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

修改一下路径

return "tcp://" + master_host, master_ports

def get_trtllm_subprocess_host_and_ports_from_env() -> Tuple[str, Tuple[str, str, str]]:
trt_subprocess_host = os.getenv("TRT_SUBPROCESS_HOST", "localhost")
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It would be better to add the "FlEXKV_" prefix

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

fixed

@axxx03 axxx03 force-pushed the feature/trtllm_adapter_tp16 branch from a5cbabe to 16d611b Compare November 27, 2025 06:42
@axxx03 axxx03 merged commit e8bf173 into taco-project:dev Nov 27, 2025
1 of 2 checks passed
peaceforeverCN added a commit that referenced this pull request Nov 28, 2025
* [bugfix] fix vllm init num_kv_heads bug

* Fix meta of non mla for trtllm_adapter (#68)

* Support TensorRT-LLM Uses FlexKV in Multi-node Tensor Parallelism (#53)

* Add change log (#69)

---------

Co-authored-by: ZHUOFAN LI <zhuofanl@nvidia.com>
Co-authored-by: Rongwei Zhang <34190091+axxx03@users.noreply.github.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants